Surface expression of the immunotherapeutic target GD2 in osteosarcoma depends on cell confluency

Abstract Background Chimeric antigen receptor (CAR) T‐cell therapy of pediatric sarcomas is challenged by the paucity of targetable cell surface antigens. A candidate target in osteosarcoma (OS) is the ganglioside GD2, but heterogeneous expression of GD2 limits its value. Aim We aimed to identify mechanisms that upregulate GD2 target expression in OS. Methods and results GD2 surface expression in OS cells, studied by flow cytometry, was found to vary both among and within individual OS cell lines. Pharmacological approaches, including inhibition of the histone methyltransferase Enhancer of Zeste Homolog 2 (EZH2) and modulation of the protein kinase C, failed to increase GD2 expression. Instead, cell confluency was found to be associated with higher GD2 expression levels both in monolayer cultures and in tumor spheroids. The sensitivity of OS cells to targeting by GD2‐specific CAR T cells was compared in an in vitro cytotoxicity assay. Higher cell confluencies enhanced the sensitivity of OS cells to GD2‐antigen specific, CAR T‐cell‐mediated in vitro cytolysis. Mechanistic studies revealed that confluency‐dependent upregulation of GD2 expression in OS cells is mediated by increased de novo biosynthesis, through a yet unknown mechanism. Conclusion Expression of GD2 in OS cell lines is highly variable and associated with increasing cell confluency in vitro. Strategies for selective upregulation of GD2 are needed to enable effective therapeutic targeting of this antigen in OS.

addition of a biological agent, liposomal muramyl tripeptide phosphatidyl ethanolamine (L-MTP-PE), yielded an increase of survival in patients with non-metastatic disease, 3 but had no benefit in metastatic OS. 4 Recurrences typically occur in the lungs, with dismal outcome despite repeated surgery. 5 Novel therapeutic approaches are needed to eliminate (micro)metastatic disease and prevent relapse.
Cellular immunotherapy with chimeric antigen receptor (CAR) engineered T cells has shown striking efficacy against refractory B-cell cancers. 6,7 Whereas the limited clinical consequences of on-target depletion of normal B cells allow to target B lineage markers, solid tumors lack surface antigens exclusively expressed on tumor cells and not on indispensable normal cells. A candidate in OS is the disialoganglioside antigen G D2 . G D2 is abundantly expressed on immature neuroectodermal tissues during embryogenesis, whereas postnatal expression is low and restricted to neuronal and mesenchymal stromal cells (reviewed in Reference 8). G D2 was found to be a safe therapeutic target for antibodies and CAR T cells in neuroblastoma, where it is abundantly expressed. 9,10 Immunohistochemistry studies have found aberrant expression of G D2 also in proportions of patients with Ewing sarcoma 11 and OS, [12][13][14] with preserved expression at recurrence. 15 But in contrast to neuroblastoma, G D2 expression in sarcomas is heterogeneous among patients and within individual tumors. To avoid antigen-negative escape, G D2 -specific immunotherapy in these cancers will have to be combined with strategies that upregulate target expression to homogeneous levels.
In previous studies in Ewing sarcoma, our group has shown that inhibitors of the histone methyltransferase Enhancer of Zeste Homolog 2 (EZH2) upregulate G D2 expression, associated with the reversal of silencing of genes encoding for enzymes in G D2 biosynthesis, effectively sensitizing antigen-negative/low tumor cells to G D2 -targeted cell therapy. 16 Overexpression of EZH2 was reported also in OS where it is associated with a highly aggressive tumor phenotype and poorer prognosis. 17,18 Here, we investigated strategies to upregulate G D2 also in OS, starting with the hypothesis that epigenetic modification by inhibition of EZH2 could induce GD2 expression also in this cancer.

| Cell lines
All OS cell lines were purchased from ATCC and the early passages after receiving were expanded and frozen in batches. The identity of the cell lines was confirmed by short tandem repeat (STR) profiling directly before freezing and the cells used for the experiments were cultured for a maximum of six passages after thawing. Tumor cells were cultured in uncoated tissue culture flasks in RPMI 1640 medium (Invitrogen, Germany) supplemented with 10% heat-inactivated fetal calf serum (FCS; Thermo Scientific, Waltham, Massachusetts) and 2 mM L-glutamine (Sigma-Aldrich, St. Louis, Missouri) at 37 C and 5% CO 2 . One hundred U/mL penicillin and 100 μg/mL streptomycin (Thermo Scientific, Waltham, Massachusetts) were added during longterm assays. The medium was changed every 3 to 4 days. Adherent sarcoma cells were harvested by trypsinization. The assays were performed by experienced individuals throughout the course of the study. The study was performed using established laboratory protocols covering the processing, freezing, storage, and thawing of cells as well as the staining procedure, data acquisition, and gating strategy.
Raw data can be provided per request.

| Flow cytometry analysis
For the analysis of G D2 expression, 100,000 tumor cells were stained with phycoerythrin (PE)-conjugated monoclonal antibody (mAb) against G D2 (14.G2a) or the corresponding PE-labeled isotype anti-IgG2a (both BioLegend, Germany). Dead cells were excluded from analysis by additional staining with Zombie Violet (Bio-Legend, Germany). Samples were fixed with 1% paraformaldehyde (PFA) and acquired directly or not later than 24 hours after staining. For each sample, 10,000 cells within the respective gates were analyzed with FACS Diva 8.0 using FACS Celesta flow cytometer (BD Biosciences, Germany) and FlowJo version 10 (FlowJo, USA). Relative fluorescence intensities (RFI) were calculated by dividing median fluorescence intensities of mAb-stained cells by those obtained with isotype antibodies (IgG2a): RFI=median GD2 /median isotype .

| Treatment with EZH2 inhibitor
OS cells were harvested and seeded in uncoated six-well plates (Sarstedt, Germany) at 0.1 to 0.5 × 10 6 cells/well in a total volume of 2 mL. After 2 hours, the EZH2 inhibitor tazemetostat (Cayman Chemicals, Ann Arbor, Michigan) dissolved in DMSO or DMSO alone as control was added at a concentration of 1, 12, 30 (Saos-2) or 60 μM (HOS), respectively. After 3 to 4 days of incubation at 37 C and 5% CO 2 , the medium was changed and the EZH2 inhibitor was added again at the same concentration. Every 7 days, the cells were harvested and analyzed for G D2 expression as above.

| Tumor spheroids
Boiled-up 1% agarose at 45 μL/well were pipetted into the wells of a flat-bottom 96-well plate (Thermo Scientific, Waltham, Massachusetts). After solidification, low-confluent MG-63 cells were harvested from monolayer cultures and 5000 cells were seeded to each agarosecoated well in a volume of 150 μL medium. The plates were incubated at 37 C and 5% CO 2 and 100 μL of fresh medium was added on day 4. Spheroids were carefully harvested using truncated pipette tips and pooled, then trypsinated and filtrated through a cell strainer (Corning, USA). Finally, cells were stained and analyzed for G D2 expression.

| Treatment with brefeldin A
The OS cell lines U-2 OS and MG-63 were incubated at 37 C and 5% CO 2 . Twenty-four hrs before reaching 50% or 100% confluency, respectively, 5 mg/mL brefeldin A (BFA) (Sigma-Aldrich, Germany) or DMSO as control were added. After 24 hours of incubation, cells at 50% or 100% confluency were harvested and analyzed for G D2 expression by flow cytometry.
After 10 hours, cells were harvested and reseeded in six-well plates at counts of 2 × 10 5 /well or 2 × 10 6 /well, respectively, to establish low-and high-confluent cell cultures. After 38 hours of incubation, cells were harvested and GFP expression was assessed by flow cytometry.

| CAR constructs and transduction of human T cells
The CAR gene GD2-BBζ and the production of recombinant retrovirus for transduction of T cells were previously described. 11,20,21 Expansion and transduction of T cells from peripheral blood were performed as described. 21 2.10 | Cytotoxicity assay   Figure S1). Even at the individual IC 30 values, tazemetostat only slightly enhanced G D2 expression in these OS cell lines within 14 days ( Figure 1B). In conclusion, in contrast to Ewing sarcoma, pretreatment with EZH2 inhibitors is not effective to overcome low and heterogeneous G D2 surface expression in OS.

| G D2 expression in OS depends on cell confluency
While screening OS cell lines for G D2 expression, we noticed that expression levels vary within individual cell lines during cell culture.
To test the hypothesis that G D2 expression levels vary with cell con-  Figure 2B). Thus, cell density is associated with G D2 surface expression not only in monolayer cultures, but also in threedimensional tumor spheroids mimicking micrometastatic tumor growth.  Figure 3A). We conclude that additional surface G D2 in confluent OS cells originates from the Golgi as a product of de novo synthesis. To understand the mechanism in more detail, we next studied individual steps along the synthetic pathway of G D2 in OS for confluency-dependent effects.  Figure S2B). To investigate whether the GD3S 5 0 UTR in OS differentially regulates translation dependent on cell confluency, we transfected MG-63 OS cells with a bicistronic vector containing the reporter gene GFP downstream of the GD3S 5 0 UTR (pCAT-5 0 UTR-GFP) to indicate 5 0 -driven translational activity, or with pCAT-GFP as control, then analyzed GFP expression at 50% and 100% confluency by flow cytometry. Insertion of the GD3S 5 0 UTR indeed noticeably decreases GFP expression, but the translational block imposed by the 5 0 UTR is not relieved by cell confluency ( Figure S2C). Thus, the GD3S 5 0 UTR does not contribute to confluency-dependent dynamics of GD3S and therewith G D2 expression in MG-63 OS cells.

| PKC stimulation does not affect G D2 expression in OS cells
The dynamic expression of G D2 in OS cells suggests a fast-acting regulatory mechanism. One candidate is activation of the PKC, which can accelerate vesicular transport of antigens from the Golgi to the plasma membrane, thereby increasing antigen surface expression. 25

| Confluency affects in vitro cytolysis of OS cells by G D2 -specific CAR T cells
Toward our goal of using G D2 as an immunotherapeutic target antigen in OS, we investigated whether confluency-dependent variation in G D2 expression affects in vitro cytolysis of OS cells by G D2 -redirected T cells. Human T cells were gene-modified to express the G D2 -specific CAR GD2-BBζ 27 and co-cultured with HOS, U-2 OS, and MG-63 target cells at 50% or 100% confluencies, respectively. High cell confluencies significantly enhanced CAR T-cell-mediated in vitro cytolysis of all OS cell targets in a cytotoxicity assay (Figure 4). Thus, high cell confluencies associated with increased G D2 target expression enhance the sensitivity of OS to G D2 -specific CAR T-cell therapy.

| DISCUSSION
Due to its restricted tissue expression, the disialoganglioside G D2 is an attractive target for cancer immunotherapy. 28 G D2 -specific monoclonal antibodies are approved for the treatment of high-risk F I G U R E 4 In vitro cytolysis of OS cells by G D2 -specific CAR T cells depends on cell confluency. Cytolysis of HOS, U-2 OS, and MG-63 cells at 50% or 100% confluency after 4 hours of coincubation with GD2-BBζ-transduced T cells. Statistical analysis by paired t-test for all E:T ratios neuroblastoma, 10,29 a cancer with abundant and consistent G D2 expression. In addition, G D2 -specific CAR T cells are starting to show clinical potential in this cancer. 9 Extending the impact of G D2 -targeted therapies beyond neuroblastoma is challenged by low levels and heterogeneity of G D2 expression in other malignancies. Consistent with the literature, 12,15,30,31 we found G D2 surface expression in the majority of OS cells lines, but levels were highly variable.
We previously reported evidence that biosynthesis of G D2 in Ewing sarcoma underlies epigenetic regulation involving EZH2, the catalytic component of the Polycomb Repressor Complex 2 (PCR2). 16 Our new finding that EZH2 is not a major regulator of G D2 expression in OS was not unexpected: Whereas Ewing sarcoma is driven by a disease-defining translocation, with consistent high-level EZH2 expression as a direct consequence of the resulting fusion protein, 32 OS is characterized by a disorganized genome with highly variable and complex chromosomal alterations. 33 Even though EZH2 can be overexpressed in OS, 17,18 loss-of-function of PRC2 was reported in OS cell lines, including HOS and U-2 OS. 17 Thus, using EZH2 inhibitors for sensitizing cancer cells to G D2 -targeted therapy may be a valuable option in Ewing sarcoma, but not in OS. Our attempts at pharmacologic upregulation of G D2 in OS by the use of PKC modulators, effective to enhance expression of CD22 in B-cell malignancies 25,34 and gangliosides in neuroblastoma, 26 were also unsuccessful.
We observed that G D2 in OS varies not only among, but also within individual OS cell lines, and that expression levels correlate with cell confluency, both in monolayer cultures and in tumor spheroids. Moreover, we found that confluency-dependent upregulation of G D2 in OS cells is mediated by increased de novo synthesis in the Golgi apparatus. We further show that cell confluency can induce expression of the key enzyme in G D2 biosynthesis, GD3S, whereas regulatory elements in the 5 0 UTR of the GD3S gene or activation of PKC are not affected. We still have to unravel the detailed mechanisms by which higher confluency in OS cells induces expression of GD3S and ultimately G D2 . Moreover, since a clear association between cell confluency and GD3S expression was shown only in one of four cell lines, additional and alternative mechanisms how cell confluency affects G D2 surface expression, for example, by enhancing transport to the cell surface or reducing degradation to less complex gangliosides, must also be considered. G D2 upregulation could be part of a cell response to metabolic stress caused by limited availability of nutrients and oxygen in confluent cell cultures and growing tumors. 35 Indeed, microenvironmental stress factors, such as hypoxia, low nutrient availability or drug exposure, can induce epigenetic remodeling associated with extensive phenotypic changes. 36,37 More specifically, cellular hypoxia can induce expression of GD3S 38 and also Sialin, 39 a sialic acid transporter, thereby enhancing expression of sialogangliosides in tumor cells.
Finally, oxidative stress caused by nutrient deprivation was found to induce expression of G D2 in breast cancer cells, 40,41 concomitant with a cancer stem cell-like phenotype. 42 Overall, it is a common observation that ganglioside expression patterns in tumor cells vary with environmental conditions, as mimicked in vitro by cellular confluency.
Translating confluency-related target upregulation into an in vivo strategy is likely to be challenging due to its multifactorial origin, limiting its clinical potential.
To what extent cell density-dependent regulation of G D2 will limit the efficacy of G D2 -targeted immunotherapy, for example, by escape of single disseminated tumor cells, remains speculative. Only clinical studies can assess the potential of G D2 as a target antigen in OS. Anti-G D2 antibody was found to have no significant efficacy against OS in a phase II trial. 43 Several G D2 -specific CAR T-cell trials that include OS patients are ongoing (listed in Reference 44). By demonstrating that optimal G D2 -CAR T-cell-mediated cytotoxicity in OS depends on cell confluency, our data support the need of strategies for overcoming heterogeneous expression of this target. An alternative means to counteract resistance of tumor cells with low antigen expression is to lower the threshold for CAR T-cell signaling and activation by modulating inherent signaling domains. 45 Tuning the reactivity of G D2 -specific CAR T cells must be weighed against potential on-target toxicities, for example, on neuronal cells with low-level G D2 expression, and against a risk for tonic T-cell stimulation triggering rapid exhaustion. 46